<div dir="auto">Reminder </div><div><br></div><div><br><div class="gmail_quote gmail_quote_container"><div dir="ltr" class="gmail_attr">On Fri, Jan 10, 2025 at 6:04 PM Emiliano De Cristofaro <<a href="mailto:emilianodc@cs.ucr.edu">emilianodc@cs.ucr.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div>Dear LOGOrithms,</div><div><br></div><div>First of all, Happy New Year! To start the year strong, we will have a talk by Yue on Friday, January 17th, at 2pm in WCH 203. Please see details below.</div><div><br></div><div>If you can't join in person, here's the <a href="https://ucr.zoom.us/j/98664053204?pwd=quPIPmylgJjHap4VkzPnaaVKk1ndi6.1" target="_blank">Zoom link</a>.</div><div><br></div><div>Cheers,</div><div>Emiliano</div><div><br></div><div><br></div><div>TITLE</div><div>Exploring Political, Social, and Cultural Biases in Pretrained Language Models<br><br></div><div>ABSTRACT<br>In this talk, I will discuss two interesting papers in NLP that examine the political, social, and cultural biases inherent in pretrained large language models. These biases raise important questions about potential harms and safety implications, which I hope to brainstorm with the group. The focus will be on the following works:<br>1. "From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models" (ACL 2023 Best Paper)<br>2. "Whose Opinions Do Language Models Reflect?" (Santurkar, Shibani, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, and Tatsunori Hashimoto. ICML 2023)<br><br>BIO<br>Yue Dong (<a href="http://yuedong.us" target="_blank">yuedong.us</a>) is an Assistant Professor in the Department of Computer Science and Engineering at the University of California, Riverside. Her primary research interests lie in trustworthy NLP, with a focus on LLM safety, hallucination reduction, and AI detection in conditional text generation tasks such as summarization. Her recent work targets red teaming of Large Language Models (LLMs) through adversarial attacks, safety alignment, and in-context vulnerabilities across generative models, including LLMs, Vision-Language Models (VLMs), and Stable Diffusion. Her recent research on VLM adversarial attacks earned the Best Paper Award at the SoCal NLP Symposium and was spotlighted at ICLR 2024. Additionally, she has served as a Senior Area Chair for top-tier NLP conferences such as NAACL 2025 & 2024, and EMNLP 2024, as well as Area Chair for ICLR 2025, ACL 2024 & 2023 and EMNLP 2022 & 2023. She has also co-organized workshops and tutorials at prestigious conferences, including EMNLP 2021 and 2023 (Summarization), NeurIPS 2021–2024 (LLM Efficiency), and NAACL 2022 (Text editing).<br clear="all"></div><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><br></div><div></div></div></div></div></div>
</blockquote></div></div>